So, I made an AI for FA

[Softly]

TLDR here's some code ----> https://github.com/HardlySoftly/Forged-Alliance-AI
(and wiki http://wiki.faforever.com/index.php?title=AI_Modding)

Its well known that AIs are generally pretty poor practice when trying to move up the ladder ranks (though they are useful in some cases).

I was wondering how hard it is to write an AI of your own, so that you can work on making it a bit more realistic to play against, and turns out its fairly easy to get something simple up and running.

I've got a little bit of code that I've shared in the hope that some more people might take a look at writing an AI as well, then we can do cool stuff like using AIs as a map tutorial rather than a replay, or AI tournaments and the like. Maybe this is something we can look at next time we do FAF Dev School.

Since I wanted the code really simple for this first release, all my AI does right now is spam t1 land and mass stall a bit. Suggestions very welcome on what I should get it to do next; is there a particular map people think I should get it to play well on, or some particular strategies you think I should teach it first?

[Heaven]

TLDR here's some code ----> https://github.com/HardlySoftly/Forged-Alliance-AI
some particular strategies you think I should teach it first?

early mantis raid into second air low-pgen/tree reclaim-based jester, then some t1 micro thing until fast t2 ghetto snipes

[Exotic_Retard]

make it do loya/harb/ras rush -> instant 2k rating

however if you teach it to beetle snipe we would just end up with tokyto : |

[Aulex]

It would be interesting if we could employ one of the various machine learning techniques to try and create a better ai

[Astrofoo]

Instead of telling it how to do a particular "strategy", make it smart enough to figure out its own. Maybe we'll learn something from it and it will revolutionize our playstyles. xD

[BRNKoINSANITY]

I love how this went from "simple AI that spams and stalls a bit" to "lets imitate humans playing" to "lets fool about with machine learning to create a warfaring AI that will teach us how to REALLY play" in less than a page of posts.

[Aulex]

I love how this went from "simple AI that spams and stalls a bit" to "lets imitate humans playing" to "lets fool about with machine learning to create a warfaring AI that will teach us how to REALLY play" in less than a page of posts.

I'm sorry, what? How does this post contribute at all?

[nine2]

Its amusing

[Aulex]

Its amusing

It's a waste of space

To follow up on my machine learning remark, something as simple as reinforcement learning could be used, the only question is what is available to the AI brain to help train the model.

[nine2]

I think you're being a little sensitive but whatever.

Machine learning is a cool idea.. Just analyze the replay vault.

But that to me sounds way too ambitious...